Analysis in Nutrition Research: Principles of Statistical Methodology and Interpretation of the Results

Analysis in Nutrition Research

Principles of Statistical Methodology and Interpretation of the Results

Editor

George Pounis

Nutrition Consultant in the Greek Food Industry, Athens, Greece

Table of Contents

Cover image

Title page

Copyright

List of Contributors

Preface

Acknowledgments

Part 1. Analysis in Nutrition Research

Chapter 1. Design of Observational Nutrition Studies

1.1. Introduction to Observational Nutrition Studies

1.2. Ecological Nutrition Studies

1.3. Cross-Sectional Nutrition Studies

1.4. Case–Control Nutrition Studies

1.5. Cohort Nutrition Studies

Chapter 2. Study Design in Experimental Settings

2.1. Introduction

2.2. Experimental Designs

2.3. Planning a Dietary Intervention

2.4. Conducting the Dietary Intervention

2.5. Conclusions

Chapter 3. Collection and Management of Dietary Data

3.1. Introduction

3.2. Dietary Information

3.3. Data Management and Dietary Analysis

3.4. Data Manipulation

Chapter 4. Dietary Pattern Analysis

4.1. Introduction

4.2. Types of Dietary Pattern Analysis

4.3. A Priori Dietary Pattern Analysis

4.4. A Posteriori Dietary Pattern Analysis

4.5. Hybrid Dietary Pattern Analysis

4.6. Challenges in Dietary Pattern Analysis

Chapter 5. Statistical Analysis of Retrospective Health and Nutrition Data

5.1. Introduction

5.2. Hypothesis Testing

5.3. Descriptive Statistics

5.4. Assessment of Normality

5.5. Confidence Interval

5.6. Pearson Chi-Square Test

5.7. Statistical Tests for Comparison of Means

5.8. Pearson Correlation Coefficient

5.9. Nonparametric Tests

5.10. Linear Regression Analysis

5.11. Logistic Regression Analysis

Chapter 6. Statistical Analysis of Prospective Health and Nutrition Data

6.1. Introduction

6.2. Descriptive Statistics

6.3. Measures to Calculate the Occurrence of a Health Outcome

6.4. Survival Analysis

Chapter 7. Meta-analysis of Nutrition Studies

7.1. Introduction

7.2. Methodology of Meta-Analysis in Nutrition Research

7.3. Statistical Methodologies Applied in Meta-Analysis of Nutrition Studies

7.4. Presentation and Interpretation of Results

7.5. Limitations and Biases

Chapter 8. Principles of Research Publication

8.1. Introduction

8.2. Citation Impact and Metrics

8.3. Article Elements

8.4. Web Tools for Writing

8.5. Reporting Data and Analysis

8.6. Publishing Process

8.7. Authorship Criteria and Acknowledgments

8.8. Strengthening the Reporting of Observational Studies in Epidemiology and Consolidated Standards of Reporting Trials Statements

8.9. Conclusions

Part 2. Challenges in Nutrition Science

Chapter 9. Mediterranean Diet: A Health-Protective Dietary Pattern for Modern Times

9.1. Introduction

9.2. Scientific Definition of Mediterranean Diet

9.3. Evidence on Health Benefits of the Mediterranean Diet

9.4. Food Components of Mediterranean Diet

9.5. Mediterranean Diet Adherence in Modern Times

9.6. Shifting to the Mediterranean Diet in the Modern Context

9.7. Conclusions

Chapter 10. Polyphenol-Rich Diets in Cardiovascular Disease Prevention

10.1. Introduction

10.2. Polyphenols: Classes, Structures, and Chemical Properties

10.3. Polyphenols in Human Diet

10.4. Dietary Polyphenols and Cardiovascular Disease

10.5. Challenges in Polyphenol Research

10.6. Conclusion

Chapter 11. Hydration and Health

11.1. Introduction

11.2. Hydration and Kidney Health

11.3. Hydration and Glucose Regulation

11.4. Implications of Fluid Balance and Obesity

11.5. Hydration and Cardiovascular Health

11.6. Hydration and Oxidative Stress

11.7. Hydration and Risk of Bladder Cancer

11.8. Conclusion

Chapter 12. Diet, Healthy Aging, and Cognitive Function

12.1. Introduction

12.2. Definition and Epidemiology of Cognitive Decline

12.3. Bioactive Components of a Healthy Diet and Cognitive Decline

12.4. Dietary Patterns and Cognitive Decline

12.5. Modern Approaches in the Study of Diet and Cognitive Health

12.6. Challenges

12.7. Conclusion

Chapter 13. Diet and Bone Health

13.1. Introduction

13.2. Brief Overview of Bone Physiology

13.3. Calcium and Vitamin D

13.4. Life Course Perspective on Nutrition and Bone

13.5. Challenges in Nutrition and Bone Research

Chapter 14. Diet and Lung Health

14.1. Introduction

14.2. Diet and Pulmonary Function

14.3. Diet and Asthma

14.4. Diet and Chronic Obstructive Pulmonary Disease

14.5. Diet and Lower Respiratory Tract Infections

14.6. Diet and Tuberculosis

14.7. Diet and Lung Cancer

14.8. Diet and Cystic Fibrosis

14.9. Diet and Interstitial Lung Diseases

14.10. Maternal Diet in Early Life and Lung Health

14.11. Challenges in Diet and Lung Health Research

Index

Copyright

Academic Press is an imprint of Elsevier

125 London Wall, London EC2Y 5AS, United Kingdom

525 B Street, Suite 1650, San Diego, CA 92101, United States

50 Hampshire Street, 5th Floor, Cambridge, MA 02139, United States

The Boulevard, Langford Lane, Kidlington, Oxford OX5 1GB, United Kingdom

Copyright © 2019 Elsevier Inc. All rights reserved.

No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions.

This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein).

Notices

Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary.

Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility.

To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein.

Library of Congress Cataloging-in-Publication Data

A catalog record for this book is available from the Library of Congress

British Library Cataloguing-in-Publication Data

A catalogue record for this book is available from the British Library

ISBN: 978-0-12-814556-2

For information on all Academic Press publications visit our website at https://www.elsevier.com/books-and-journals

Publisher: Charlotte Cockle

Acquisition Editor: Megan Ball

Editorial Project Manager: Michelle Kublis

Production Project Manager: Paul Prasad Chandramohan

Cover Designer: Christian Bilbow

Typeset by TNQ Technologies

List of Contributors

Claudia Agnoli,     Epidemiology and Prevention Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy

Krasimira Aleksandrova,     Nutrition, Immunity and Metabolism Senior Scientist Group, Department of Nutrition and gerontology, German Institute of Human Nutrition Potsdam-Rehbruecke (DIfE), Nuthetal, Germany

Daria C. Boffito,     Department of Chemical Engineering, Polytechnique Montréal, Montréal, QC, Canada

Emmanouil Bouras,     Department of Medicine, Aristotle University of Thessaloniki, Thessaloniki, Greece

Ock K. Chun,     Department of Nutritional Sciences, University of Connecticut, Storrs, CT, United States

Michel de Lorgeril,     Laboratoire Cœur et Nutrition, TIMC-IMAG, School of Medicine, University of Grenoble-Alpes, Grenoble, France

Elaine Dennison,     MRC Lifecourse Epidemiology Unit University of Southampton, United Kingdom

Romina di Giuseppe,     Institute of Epidemiology, University Kiel, Kiel, Germany

Monica Dinu,     Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy

Pauline M. Emmett,     Centre for Academic Child Health, Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom

Konstantinos I. Gourgoulianis,     Respiratory Medicine Department, University of Thessaly, School of Medicine, Larissa, Greece

Anna-Bettina Haidich,     Department of Medicine, Aristotle University of Thessaloniki, Thessaloniki, Greece

Lisa T. Jansen,     University of Arkansas, Fayetteville, AR, United States

Louise R. Jones,     Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom

Stavros A. Kavouras,     Hydration Science Lab, Arizona State University, Phoenix, AZ, United States

Themis Koutsioukis,     Respiratory Medicine Department, University of Thessaly, School of Medicine, Larissa, Greece

Vittorio Krogh,     Epidemiology and Prevention Unit, Fondazione IRCCS Istituto Nazionale dei Tumori, Milan, Italy

Sang Gil Lee,     Department of Food Science and Nutrition, Pukyong National University, Busan, Korea

Foteini Malli

Respiratory Medicine Department, University of Thessaly, School of Medicine, Larissa, Greece

Technological Institute of Thessaly, Nursing Department, Larissa, Greece

Kate Maslin,     MRC Lifecourse Epidemiology Unit University of Southampton, United Kingdom

Dimitra Mastorakou,     Leicester, United Kingdom

Melissa Melough,     Department of Nutritional Sciences, University of Connecticut, Storrs, CT, United States

Kate Northstone,     Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom

Gregory S. Patience,     Department of Chemical Engineering, Polytechnique Montréal, Montréal, QC, Canada

Paul A. Patience,     Department of Electrical Engineering, Polytechnique Montréal, Montréal, QC, Canada

George Pounis,     Alimos, Athens, Greece

Mikael Rabaeus,     Geneva, Switzerland

Junichi Sakaki,     Department of Nutritional Sciences, University of Connecticut, Storrs, CT, United States

Patricia Salen,     Laboratoire Cœur et Nutrition, TIMC-IMAG, School of Medicine, University of Grenoble-Alpes, Grenoble, France

Adam D. Seal

University of Arkansas, Fayetteville, AR, United States

Hydration Science Lab, Arizona State University, Phoenix, AZ, United States

Francesco Sofi

Department of Experimental and Clinical Medicine, University of Florence, Florence, Italy

Unit of Clinical Nutrition, Careggi University Hospital, Florence, Italy

Don Carlo Gnocchi Foundation, Onlus IRCCS, Florence, Italy

Hyun-Gyu Suh

University of Arkansas, Fayetteville, AR, United States

Hydration Science Lab, Arizona State University, Phoenix, AZ, United States

LynnDee G. Summers,     University of Arkansas, Fayetteville, AR, United States

Caroline M. Taylor,     Centre for Academic Child Health, Population Health Sciences, Bristol Medical School, University of Bristol, Bristol, United Kingdom

Konstantinos K. Tsilidis,     University of Ioannina School of Medicine, Ioannina, Greece

Preface

Analysis in Nutrition Research: Principles of Statistical Methodology and Interpretation of the Results describes, in a comprehensive manner, the methodologies of the quantitative analysis of data originating specifically from nutrition studies. The book summarizes various study designs in nutrition research, research hypotheses, the proper management of dietary data, and analytical methodologies, with a specific focus on how to interpret the results of any given study. In addition, it provides a comprehensive overview of the methodologies used in study design and the management and analysis of collected data, paying particular attention to all of the available, modern methodologies and techniques.

Readers will find an overview of the recent challenges and debates in the field of nutrition research that will define major hypotheses for research in the next 10 years. Nutrition scientists, researchers, and undergraduate and postgraduate students will benefit from this thorough publication on the topic.

Acknowledgments

It is a pleasure to thank all of the people who made this title possible.

I am grateful to the contributors who kindly participated in preparing the various chapters of this book and shared with us their knowledge, experience, and expertise.

I would not have achieved this ambitious goal without the support of Elsevier and I wish to thank Megan Ball, who trusted me with the editorial role.

I want to extend my sincere gratitude to my colleagues from Elsevier, who supported me in the various steps of this publication.

My final thoughts go to my family and Eleftheria. I would not have gone so far without their love and encouragement.

Georgios

Part 1

Analysis in Nutrition Research

Outline

Chapter 1. Design of Observational Nutrition Studies

Chapter 2. Study Design in Experimental Settings

Chapter 3. Collection and Management of Dietary Data

Chapter 4. Dietary Pattern Analysis

Chapter 5. Statistical Analysis of Retrospective Health and Nutrition Data

Chapter 6. Statistical Analysis of Prospective Health and Nutrition Data

Chapter 7. Meta-analysis of Nutrition Studies

Chapter 8. Principles of Research Publication

Chapter 1

Design of Observational Nutrition Studies

George Pounis     Alimos, Athens, Greece

Abstract

Systematic observation along with accurate and systematic measurement has been the basis for the development of nutrition science throughout the years. A considerable proportion of dietary risk factors for various diseases have been identified and first studied through observational nutrition studies. In this chapter, the reader will have the opportunity to be guided in the design of ecological, cross-sectional, case–control and cohort nutrition studies. The basic concepts, methodologies, and the challenges for investigators will be discussed along with appropriate survey examples from the literature.

Keywords

Case–control study; Cohort study; Confounder; Cross-sectional study; Diet; Ecological study; Epidemiology; Incidence; Nutrition; Nutrition research; Observational nutrition study; Prevalence; Prospective study; Retrospective study; Sampling; Study design

Chapter Outline

1.1 Introduction to Observational Nutrition Studies

1.2 Ecological Nutrition Studies

1.2.1 Description

1.2.2 Challenges

1.2.3 Example of Ecological Study on Diet and Cancer

1.3 Cross-Sectional Nutrition Studies

1.3.1 Description

1.3.2 Sampling

1.3.3 Challenges

1.3.4 Example of the National Health and Nutrition Examination Survey

1.4 Case–Control Nutrition Studies

1.4.1 Description

1.4.2 Sampling

1.4.3 Challenges

1.4.4 An Example in the Study of Hodgkin Lymphoma

1.5 Cohort Nutrition Studies

1.5.1 Description

1.5.2 Sampling

1.5.3 Challenges

1.5.4 Example of the Nurses' Health Study

References

1.1. Introduction to Observational Nutrition Studies

The science of human nutrition that is frequently described as nutrition science or nutrition is the science of food, the nutrients and other substances therein, their action, interaction and balance in relation to health and disease, and the processes by which the human organism ingests, absorbs, transports, utilizes and excretes food substances [1,2].

Systematic observation along with accurate and systematic measurement has been the basis for the development of science in various fields from mathematics and physics to sociology and medicine. Although nutrition science is relatively young compared with others, it has been firmly tied to observation and the measurement since the mid-1950s, when the first evidence-based conclusions regarding the association of nutrition and health were determined [1–13].

As part of nutrition science, nutritional epidemiology incorporates valuable information related to the methodologies and techniques of observation and measurement applied to generate evidence-based conclusions regarding the interaction between diet and health, among others. A considerable proportion of dietary risk factors for various diseases have been identified and studied through epidemiological health and nutrition studies.

An observational nutrition study may be defined as the detailed investigation and analysis of information provided by a retrospective or prospective systematic observation and measurement of a sample's dietary factors (exposures) and health characteristics (outcomes), in which the researcher does not willingly influence the collected information.

Observational nutrition studies have an epidemiological character and aim to generalize conclusions derived from the investigation and analysis of sample data from the reference population [6,14]. They are often named population-based studies. According to the nature of the collected data, they are divided into retrospective and prospective studies [6,14,15]. Types of retrospective surveys include ecological, cross-sectional, and case–control studies whereas prospective surveys include follow-up and longitudinal studies, which are generally labeled cohort studies.

All of these study designs provide valuable descriptive information about adherence to dietary patterns, the consumption of foods and nutrients, the presence of certain dietary behaviors, and other dietary characteristics in a population base. In addition, they aim to test hypotheses related to the association of dietary exposures with health outcomes. Although by themselves these studies are not enough to prove a cause-and-effect relationship between a dietary factor and a health outcome, associations observed mainly in cohort surveys assign the potential existence of causality. In these cases, further investigation through intervention trials and meta-analysis is essential when it is ethically and technologically feasible.

Evidence-based nutrition science or practice is well-promoted as the preferable and most accurate methodology to make decisions in all related disciplines; it aims to maintain or improve the health of individuals, groups, and populations [10]. Observational nutrition studies are a first and important step in the evidence-based concluding chain of nutrition science.

All of the various types of observational nutrition studies have certain advantages and limitations that derive from their nature, and which should be considered throughout the stages of the survey (i.e., design, implementation, analysis, presentation, and publication of results and conclusions). In addition, the methodology of the collection, the management and statistical analysis of data, and the presentation and interpretation of the results vary among the different designs. These issues are addressed in other chapters of the book (i.e., Chapters 3–8).

The following sections offer an overview of the various types of observational nutrition studies illustrating their basic concepts and design methodologies, providing survey examples from the literature.

1.2. Ecological Nutrition Studies

1.2.1. Description

An ecological nutrition study is the first level of systematic and comparative observation and measurement of the dietary characteristics of large populations (mainly geographically oriented), usually in parallel with the systematic and comparative observation and measurement of health-related indices in the same populations [6]. The experience of variations in an index of health status between populations with variations in the average value of a dietary factor introduces the concept of the potential association between the dietary exposure and the health outcome. This remains to be confirmed in other observational and experimental settings.

Implementation of this type of survey requires at least that a population-based measure of a dietary factor and an index of health status be available for two or more populations. For instance, a country with 20% per capita intake of calories from fats has a lower incidence of colon cancer than does a country with 45% per capita intake of calories from fats. This introduces the concept that the consumption of fat in these populations may have a role in explaining the variation presented in their incidences of colon cancer.

Although the major aim of ecological studies is to characterize populations over a dietary factor and an index of health status rather than evaluating the association between a dietary exposure and a health outcome [14]. More often, this survey type is used to explore geographical differences in the diet and health of large populations. It is also used to compare changes in diet and the health status of populations over time. Sometimes an ecological study is the only design that can be planned if the dietary data of the study population are unavailable at an individual level.

An advantage of this survey type is that the average population-based diets and the per capita consumption of food tend to be stable over time [14]. Moreover, the indices of diseases are mainly derived from large samples and are under limited biases. For these reasons, it can be hypothesized that variations in health indices and dietary factors in the studied populations may have longer time of pre-occurrence.

However, a serious limitation is that the same variations can also be attributed to determinants of disease other than dietary, such as genetic, environmental, clinical, and lifestyle. This poses a major disadvantage of ecological studies along with the limited opportunity to reproduce results, especially for international surveys.

Despite the inability of this study design to validate a cause-and-effect relationship between a dietary exposure and a health outcome, it has been proved to be effective in generating scientific hypotheses for further observational and experimental studies.

1.2.2. Challenges

Among the major concerns for the designers of an ecological nutrition study is the identification of appropriate population-based measures of the dietary characteristics and health status of the populations under investigation.

Estimates of the population dietary intake are often retrieved from preexisting data generated by systematic measures and evaluations performed for other reasons (i.e., economic studies, census). Sources of dietary information include national and international data about the per capita consumption of foods, data from household surveys, and individual survey data from representative sample populations.

An important consideration for the survey's investigators is a study of the methodology used for data collection and management. The use harmonized national and international population-based dietary data ensures their comparability. This purpose is assisted by using common data collection and management methodologies.

Another challenge to conducting an ecological study is evaluating the level of accuracy of population-based dietary data. Biases related to the methodology of data acquisition should be limited. For instance, frequently, the data for the per capita consumption of food in a country are retrieved by economic studies for food sales, imports, and exports. Because of their aims and methodology, these surveys cannot control for the amount of food that is wasted. It may be true that a significant amount of food that has been sold to consumers was not actually consumed. The percentage of wasted food among countries with different socioeconomical standings may vary significantly. This may lead to biased conclusions regarding variations observed in food consumption and the health status of populations under investigation.

1.2.3. Example of Ecological Study on Diet and Cancer

In 1975, Armstrong and Doll published findings of their ecological study on the association of dietary factors with the incidence of various types of cancer and cancer mortality rates [16]. This was one of the first observational studies on the topic, and despite the methodological limitations derived from its design, it has been a reference work for further studies on the topic (i.e., more than 1700 citations in Scopus Metrics).

The researchers used data on the incidence rate of 27 cancer types in 23 countries for individuals aged 35–64  years, derived from the Union for International Cancer Control in 1966 and 1970 [17,18]. Data on the cancer mortality rates for 14 cancer types in 32 countries were taken from Segi et al. and the World Health Organization (1967–69, 1970) [19–21]. This information was studied in association with data on the international per capita consumption of various foods derived from various reference sources.

The results of the study were presented grouped by the type of cancer, and graphs showing related trends were used. Various correlations were noted between dietary variables and cancer incidence or mortality. The researchers pointed out in their conclusions that the most strong associations were of meat and total fat with colon, rectum, and breast cancer. Interestingly, they also stressed the limitations of their study and noted that other population characteristics might act as confounding factors in some of the observed associations.

1.3. Cross-Sectional Nutrition Studies

1.3.1. Description

Cross-sectional nutrition studies are often named descriptive surveys; they are a basic type of observational study. They mainly collect and analyze dietary and health data from a sample population at the certain time point in the participants' recruitment [22,23]. For this reason, information provided by participants is a snapshot of the sample's dietary or other characteristics at the time of recruitment.

Cross-sectional studies in nutrition science provide descriptive results for the degree of exposure to dietary factors of a sample population. They evaluate current dietary habits of participants and offer valuable information about the consumption of foods and other dietary components. Energy, water, and nutrient intake, adherence to dietary patterns, and other dietary factors can be estimated [6,24–27]. Moreover, by collecting biological samples and biobanking [28,29], markers of food and nutrient bioavailability and metabolism can be measured in the sample population base. Frequently, participants undergo anthropometric screening and the related information is collected and analyzed [30–32].

Because these studies run in parallel with other health-related studies, they provide important information for the prevalence of a disease or health condition [33]. It is crucially important that the diagnostic criteria used in the survey are the most up-to-date and in accordance with the recent literature. There are several expressions of prevalence; a simple one is:

For example, when we read that the prevalence of obesity was 25% in November 2015 in a sample population, we may understand that of the 100 participants who were screened at that specific time point, 25 were obese.

During a cross-sectional nutrition survey, the degree of exposure to various potential risk factors such as lifestyle, socioeconomic, environmental, genetic, and clinical is usually measured in parallel at the same time point of the study sample's recruitment [8,34–36]. The choice of the factor that will be assessed depends on the aims of the survey. It is highly essential that all screening tools used to assess the risk factors be valid and accurate.

The time frame of a cross-sectional study is limited to a time point, so it is impossible to identify cause-and-effect relationships between dietary exposures and health outcomes [33,37,38]. The definition of causality in an observed association requires at least that the dietary exposure occurred before the diagnosis of a disease or health condition [6,14,15]. This condition is not confirmed by a cross-sectional design.

Statistical methodologies are available to evaluate the association of a dietary exposure and a health outcome and they are also widely used by researchers, such as linear regression modeling and generalized linear modeling (see Chapter 5). However, the interpretation of significant associations in cross-sectional settings should be limited by the study's inability to express causality. When these associations are supported by literature data (i.e., in vitro or other studies) that offer biological plausibility, they may guide further hypothesis testing and studies in other survey designs (i.e., cohort studies, meta-analyses, and randomized control trials).

Weakness points of this type of observational study also relate to dietary assessment methodologies and tools, which sometimes fail to evaluate the dietary habits of participants accurately. In particular, bias related to the recall of dietary information from sample members is evident whereas potential alterations of long-term dietary habits resulting from the previous presence of a disease or a health condition cannot be thoroughly investigated. Thus, the observed statistical associations between dietary exposures and health outcomes may be biased. These issues are addressed in detail in Chapter 3, where the collection and management of dietary data are discussed.

Despite the limitations presented by the cross-sectional design, the data extracted by this type of nutrition study provide useful information for public health professionals. In fact, cross-sectional surveys offer comprehensive screening of the nutritional and health status of sample populations at certain time points. For this reason, many health institutions worldwide regularly plan, fund, conduct, analyze, and publish this type of study (related examples are presented in Section 1.3.4).

1.3.2. Sampling

Given the importance of cross-sectional nutrition surveys in public health perspectives, the need exists to generalize cross-sectional results derived by analyzing sample data to the reference population [39]. The proper choice of the reference population and the methodology for recruiting a representative sample are crucial.

The process of selecting the reference population requires a precise definition of the main hypothesis testing, dietary exposures, health outcomes, and other studied factors (Fig. 1.1). The calculation of the sample size, which will ensure the statistical power of the study's results, is the next important step [40]. There is available free or paid software that allows these calculations with a certain level of accuracy.

When the reference population and the sample size have been well-defined, a decision should be made about the sampling methodology or sample design. This is the method that will be used for the representative selection of sample members from the reference population. Sample members can be individuals, couples, households, or even schools. The finest approach is to use random sampling [15]. In this method, theoretically the probability of a member of the population being included in the study's sample is the same for all members of the population. In this scenario, possible selection biases are limited [33].

Several sample designs are available: simple random sampling, systematic sampling, multistage sampling, and stratified sampling [6,15,23]. Application of the sampling methodology depends on available resources and the decisions of the investigators regarding how the sample will best represent the reference population.

Simple random sampling is a sample design that requires an available comprehensive sample frame (i.e., a list of all potential sample members of the reference population) [41,42]. For instance, a complete sample frame for the adult population of a region can be the electoral register. Once the sample frame has been identified, a number is given to all of its members. The random selection of sample members who will be invited to participate in the study is based on a random sequence of numbers, which can be generated by software.

Figure 1.1  Sampling of cross-sectional nutrition studies.

Systematic sampling is another sample design based on the conditions that potential participants are arranged in a type of sequence and sample members are selected according to a random starting point and a fixed periodic interval [41–43]. Examples of the potential participants' sequence are a series of index cards, houses along the side of a street, and patients who arrive at a hospital or clinic. A periodic interval is commonly named sampling interval and is calculated by the following formula:

For example, if we want wanted to select a random group of 200 sample members out of 2000 members of the reference population, the sampling interval is 10. Staring from a random point of the population sequence (e.g., the 18th member), every 10th member of the sequence should be invited to participate in the study (i.e., the 18th, 28th, 38th…) until 200 sample members will be reached.

To ensure that the sample is random, the starting point should be random and the sampling interval should not correspond with any repeated pattern in the sequence of the reference population. For instance, if we have a list of couples and the women are always placed first, if we start from the third member of the list and the sampling interval is 8, the sample members will be all women. This example can be easily understood, although there are might be repeated patterns that are difficult to identify and may be associated with specific characteristics of the sample.

Multistage sampling is one of the most popular sample designs for cross-sectional nutrition studies because it is more realistic and ensures the representativeness of a sample deriving from a large geographical area [44]. Frequently, cross-sectional nutrition studies aim to evaluate the current dietary habits of large populations such as a country. It would be unrealistic to select a sample of 1000 participants from a country using a simple random sample design. The first problem would be the existence of a comprehensive sample frame. Even if this were available, the sample members indicated by random numbers might live far from each other, which would make it impractical in terms of time and resources to recruit them.

To apply multistage sampling in this case, we may divide the country into regions, and from that list we may select a random sample of regions (first-stage sampling). In each of the randomly selected regions a list of towns should be drawn and a random sample of them can be selected (second-stage sampling). In each randomly selected town, a sample frame may be available to make a random selection of sample members (third-stage sampling). Even if the sample frame is incomplete, further sampling stages can be used with geographical or other related criteria until a random, representative, and feasible sample in terms of recruitment practices can be selected.

Stratified sampling is a sample design that is used in parallel with simple or the multistage sampling [12]. A stratified random sample occurs after dividing the population into specific subgroups (i.e., strata) often according to demographic characteristics such as gender, age, and socioeconomic status. In every subgroup, a random sample is selected by applying simple or multistage sampling. This type of sample design is appropriate when there may be indications that dietary exposures or health outcomes vary between strata.

1.3.3. Challenges

A major concern of researchers regarding the final sample for all types of sample design is to achieve an optimal level of randomness and representativeness of the reference population. In this effort, a serious challenge is to identify an up-to-date and complete sample frame, because many cross-sectional nutrition studies refer to large populations. The use of random techniques in selecting sample members from a relatively complete sample frame increases the probability that the final study sample will be representative.

The census of a population may provide valuable information to recruiters about demographics and other characteristics of the reference population. During various stages of the study's sample recruitment, it is a common practice for some of the collected data to be assessed compared with census data to determine significant differences. The absence of significant variations between them indicates representativeness. Data from previous studies in the reference population could be also used for the same reason.

Another challenge for the designers of cross-sectional nutrition studies is the response rate. To calculate the response rate, we may use the following formula:

For example, from a sample frame of 2000 sample members, 250 randomly selected sample members were invited to participate in the study and 150 were finally interviewed. Thus, the response rate is 60%.

There are several reasons why a sample member might not participate in the study. The designers of a cross-sectional survey should plan techniques that deals with this issue (e.g., prompting by telephone and mail, study advertisements, letters outlining the importance of replying). All possible efforts should be made to ensure a high response rate to minimize possible selection bias.

A common dilemma regards what response rate is acceptable. There are empirical rules proposing that a response rate over 80% might set a level of confidence for the recruiters. However, consider a situation in which the response rate was 90% but 10% of nonresponders had very specific and distinct characteristics from responders that made them decide not to respond. In this example, selection bias might be present despite the high response rate.

To evaluate bias introduced by nonresponse, it is important to obtain some information from sample members who refused to participate or could not be reached. The design and application of a short questionnaire acquiring important information from a random sample of nonresponders might be useful. These data could be later compared with those from responders and help the researchers to identify significant differences. Commonly, demographics (e.g., gender, age, socioeconomic status) or other information might be available for the members of the sample frame. These data could be comparatively evaluated for responders and nonresponders.

Equally important to selection bias that is related to the response rate is biased responses deriving from the use of inappropriate or inaccurate assessment methodologies or tools. Investigators face myriad challenges when decisions need to be made regarding the use of valid methodologies and tools to measure dietary exposures, health outcomes, and other studied factors. The same is for the application of appropriate data management and statistical analysis methodologies. These challenges are evident in all study designs; they are thoroughly presented and discussed in Chapters 3–6.

Generally, compared with other observational study designs (i.e., case–control and cohort studies) or randomized control trials, cross-sectional nutrition studies are less expensive and time-consuming [6]. Moreover, in parallel with the evaluation of dietary data, valuable descriptive information on health parameters and other characteristics of the population can be assessed. The prevalence of diseases can be estimated especially for longer-lasting diseases (i.e., chronic diseases) because diseases lasting for a short period may be underestimated (i.e., the Neyman bias) [7,45]. Considering that a cross-sectional study refers to the current status of a population, it gains great value for public health planning means. Although associations between dietary exposures and health outcomes that are identified through a cross-sectional design do not have a cause-and-effect character, they guide further hypothesis testing and studies in the field.

1.3.4. Example of the National Health and Nutrition Examination Survey

The National Health and Nutrition Examination Survey (NHANES) is a unique rolling research program that collects cross-sectional information for the nutritional status and health of children and adults in the United States [46]. It was initiated in the early 1960s, and from the 1999 onward, it became a continuous program focusing on a variety of health and nutrition aspects. As a project of National Center for Health Statistics (NCHS), NHANES collects data from a representative sample of 5000 persons each year. The NCHS is part of the Centers for Disease Control and Prevention and has the responsibility of producing vital and health statistics for the nation. Participants are located in counties across the country, 15 of which are visited each year. To produce reliable statistics, NHANES oversamples persons aged 60  years and older who are African American and Hispanic.

For all participants, the NHANES data collection includes an in-person household interview that is performed by highly trained field investigators. In a second stage, all sample members are asked to complete a comprehensive health examination, which consists of clinical tests, laboratory studies, and additional interviews. These examinations are performed by experienced scientific personnel in specially designed and equipped mobile centers, which travel to locations throughout the country. The third stage includes postexamination interviews and questionnaires that are administered by telephone or mail.

Three major types of dietary data are collected in NHANES: dietary behavior, a 24-h dietary recall, and a food frequency questionnaire (FFQ). During the household interview, information is obtained on dietary behavior including topics such as dietary modifications owing to health conditions and dietary supplement use. During the in-person examination, a 24-h dietary recall is administered. These dietary interviews are conducted by dietary interviewers. Survey participants aged 12  years and older complete the dietary interview on their own. Proxy respondents report for children who are aged 5  years and younger and for other persons who cannot self-report. Proxy-assisted interviews are conducted with children aged 6–11  years. A second 24-h dietary recall is administered to all participants by telephone 3–10  days after the in-person health examination.

The collected data provide valuable information about the nutritional and health status of the participants. They are used to estimate the prevalence of various major diseases and the distribution of dietary and other risk factors in the population. In addition, the association of the nutritional status with health promotion and disease prevention is evaluated. The NHANES results and outputs set the basis for further studies and assist in developing meaningful public health policies.

1.4. Case–Control Nutrition Studies

1.4.1. Description

The case–control nutrition study is another type of observational study with retrospective character. In this survey type, participants with a newly diagnosed disease (cases) are compared with matched healthy individuals (controls) for the past exposure to dietary factors [6,14]. Observation of significant differences in the dietary exposures among groups of cases and controls suggests that the exposure may be associated with decreased or increased prevalence of the disease [47].

This study type allows an estimation of the odds ratio (OR) (see Chapter 5), which represents the odds that a health outcome will occur given a particular dietary exposure, compared with the odds of the outcome occurring in the absence of that exposure [22,23].

Study of the ORs introduces the concept of the association between dietary exposure and health outcome. However, data related to the dietary habits of participants for the period previous to the diagnosis of the disease are retrieved mainly by memory recall (using various dietary assessment methodologies) and are under recall bias [15,48,49]. In this context, interpretation of the results should be always performed with caution.

A basic characteristic of this survey type is the matching of cases and controls for some confounding factors (e.g., gender, age, socioeconomic status). Conceptually, a confounder is a variable that influences both the dietary exposure and the health outcome, causing a false association between them [33]. Matching is a kind of control for confounding effects in an observed association between a dietary exposure and a health outcome [14]. In addition, statistical methodologies that are used to analyze case–control data allow further controlling for potential confounding (see Chapter 5) [11,50].

Case–control nutrition studies are generally efficient for studying the dietary implications of rare diseases (i.e., diseases with a low incidence in the reference population that need a long induction period) [15]. They also demand less time and resources to be conducted compared with cohort studies (see Section 1.5) because they require smaller samples and no follow-up is necessary.

1.4.2. Sampling

A major aspect of sampling methodologies applied to case–control studies is the proper choice of cases and controls. A variety of approaches are available to recruit sample members for both groups that aims to ensure a certain level of representativeness of the reference populations and avoid sampling bias.

The first dilemmas for case–control study designers are related to the methodology that will be used to select the cases. The precise and accurate definition of the disease or the health condition under investigation is the first, crucial step [9]. The diagnostic criteria that will be applied should ensure the highest level of accuracy possible to identify the cases, and they should be the most up-to-date.

In the first stages of design, the recruiters should also decide on the inclusion of prevalent or incidence cases. Prevalent cases are existing cases that are present in the population during the time frame of the study's sample recruitment …. This definition includes both new and old cases. The incidence cases are the newly diagnosed individuals within the period of the sample recruitment ….

The major advantage of selecting incidence instead of prevalent cases in a case–control nutrition study is that dietary information recalled from newly diagnosed individuals tends to be more representative of past exposure to the investigated dietary factors. The dietary habits of prevalent cases may have been amended in the time between the diagnosis of the disease and the participant's recruitment, as an effort to adhere to a healthier lifestyle.

The selection of prevalent cases is sometimes the adopted methodology because it needs less time and resources. In this scenario, the check for differences in the investigated dietary exposures between groups of prevalent cases with various times of diagnosis may help in understanding the level of biases induced in the study.

The choice of incidence cases should be highlighted as the preferred methodology in case–control nutrition studies because the exposure is always a dietary factor that is sensitive to alterations after the diagnosis of a disease. Moreover, by recruiting only newly diagnosed cases and selecting controls with a methodology to be representative of the population in which the cases arise makes identification of an association between a dietary factor and a disease mimic those of a cohort study.

Cases that will be recruited into the study can be selected from the population of patients attending a hospital or a network of hospitals. If these health care units are the major points of treatment for patients of a disease under investigation, the sample members will be more representative of the diseased population.

Another methodology is the recruitment of cases from the population over a fixed period of time. In this instance, there should be evident difficulties in tracing the subjects and the refusal rate may be greater. These factors may lower the representativeness of the diseased population. There are certain situations in which this method is applied; they are related to the existence of a comprehensive registration system for patients of the disease under investigation …. This allows the identification of a complete sample frame of the diseased population.

Equally crucial as the choice of the cases is the appropriate selection of controls that fit the same eligibility criteria as cases apart from those that related to the diagnosis of the disease under investigation. The study designers should take into account any special characteristic or circumstance of the study.

A first approach is to select the controls randomly from the sample frame of the population from which the incident cases were identified [51,52]. It is a time- and resource-consuming methodology, requiring a complete sample frame that is challenging to find, and the selection of cases may be under selection bias of nonresponse.

A second methodology is to select controls from the close environment of the diseased person, such as relatives and friends [51,52]. In this instance, the level of the exposure of the controls to dietary factors may have similarities with that of the cases, leading to overmatching and making it impossible to identify significant differences. The same challenge is evident when the controls are drawn from the sample frame of the neighborhoods of the cases. People living in the same areas interact each other in a way that may lead to overmatching of cases and controls, especially for the investigated exposures (i.e., dietary habits).

Another common approach is to select controls from the same hospitals from which the cases have been drawn [51,52]. This methodology requires the controls to have been hospitalized for a reason different from the disease under investigation. Many health conditions have common soils in terms of dietary risk factors that are attributed to their development and progression. The selection of the sample members in a case–control nutrition study, however, requires that the level of dietary exposure of the cases and controls not adhere to a common pattern. In this way, the potential differences in dietary factors between cases and controls corresponding to a variation in the prevalence of the disease may be assessed.

Some selective factors may be present that bring patients to a specific hospital (i.e., social status, residence). When controls are chosen from the same hospitals where cases have been drawn, both groups become more homogeneous with respect to these selective factors, limiting their effects.

The selection of controls from a hospital base also seems to enhance the completeness and accurateness of the collected information. Controls experiencing a recent hospitalization and illness tend to be more cooperative with the study's procedures because they better appreciate and respect the significance of the aims of a health survey.

1.4.3. Challenges

During the design of a cross-sectional nutrition study, the investigators may force various challenges related to recruiting the cases and controls and the method applied to match the two groups.

The first challenge is the uncertainty presented in classifying patients into the diseased group. It is sometimes evident that the diagnostic criteria do not provide with a definite decision regarding the presence of a disease. In these instances, further classification of cases into groups of level of certainty for the presence of the disease (i.e., high, moderate, low) may assist in data analysis.

Ideally, when the definition of the cases is performed with a high level of certainty, this results in a homogeneous group of patients with a common pathology. The observation of significant associations in these studies may provide a better etiological background.

There are some occasions, especially for newly diagnosed cases, in which patients are too ill to respond or to recall information about diet or other risk factors. This generates the need for some eligibility criteria that should be decided with respect to general hypothesis testing, always considering avoiding selection bias. They should be common for cases and controls apart from those related to the diagnosis of the disease under investigation. The number of the invited eligible participants who refused to participate in the survey should also be recorded along with the reason for nonparticipation and some other information (i.e., age, gender, sociodemographic characteristics) if possible.

Another challenge for the study designers regards the methodology to be used to match the cases and controls. A first, common approach is that for every case, one or more identical controls, in terms of some characteristics (matching factors), is recruited [33]. The matching factors are confounders of the association between the dietary exposure (see Section 1.4.1) and the disease and should be carefully chosen.

Their selection should be based on strong evidence supporting their confounding role. Frequently used factors are gender, age, residence, and nationality. Researchers should avoid matching with many factors because this may cause significant similarities in terms of dietary exposures between cases and controls. Overmatching makes it difficult to study dietary exposures as potential causes for the disease or the health condition under investigation [33]. Researchers can perform further adjustments for potential confounders during statistical analysis of the data.

The second matching approach requires the group of healthy controls to be recruited with an equal proportion of some matching factors with the group of cases [15]. For instance, if in the group of cases 35% are females, 55% are at age <35  years, and 25% have a higher education degree, then the groups of control should have the same proportion in these characteristics.

The ratio of the number of controls per case should be also decided. It determines the